How It Works
We scan GitHub, npm, PyPI, and MCP registries to find every public MCP server.
For each server with a GitHub repo, we fetch real data: stars, forks, last commit, README content, test files, package configs, and open issues.
Our algorithm computes a 0-100 score across 8 parameters. Each parameter has a weighted maximum (e.g., recency = 20pts, documentation = 15pts).
Scores map to letter grades: A (90-100), B (70-89), C (50-69), D (30-49), F (0-29). We show all grades so you can filter by quality.
Scores are recalculated regularly. A server that goes stale will drop in score. Active maintenance is rewarded.
Why Trust Our Scores?
No AI hallucination. Every score is computed from real GitHub API data — not from LLM guesswork. We read your README, check your test directory, count your stars over time, and verify your last commit.
If a server doesn't have a public GitHub repo with verifiable stats, we flag it with a "⚠ no GitHub data" badge (you'll still see it listed, but we can't score internal components).